Integrating LLMs and RAG Techniques Into Mathematics Learning for Engineering Students

In this study, we present a novel pipeline designed to enhance the capabilities of large language models (LLMs) for solving mathematical problems (exercises, course) using a retrieval-augmented generation (RAG) methodology. Recog-nizing that a straightforward RAG approach is inadequate for significantly improving the reasoning abilities of LLMs, we propose an advanced method to address these limitations. Our approach integrates sophisticated retrieval techniques with generative models to provide more accurate and contextually relevant solutions to mathematical problems. By combining these elements, we aim to improve the problem-solving efficacy of LLMs, making them more effective tools for educational purposes and beyond. This research demonstrates the potential of enhanced RAG methodologies in advancing the application of LLMs in mathematics education, offering a promising pathway for overcoming existing challenges in the field.